Weighing primers testing results ANYONE DOING IT ???

[Ned Ludd]

I thought the idea of weighing primers was just a joke. But, judging from how serious some members of this forum are, it looks like I'm going to have to step up my reloading procedure.

I would think to get any meaningful results from weighing primers one would need to dig deeper than simply weighing the intact primers. I'm going to try separating the several components and weighing them individually. The weights of the anvil, the cup, and the pressure-sensitive ignition compound need to be separated because, it seems to me, that the ignition compound is the component most likely to effect group size.

In other words, if the anvils and ignition compound of a group of primers all weigh the same and only the cup weight varies, then I suspect performance differences would be small. On the other hand, if all the anvils and cups are identical, but there are significant variations in the weight (amount) of the ignition compound, then it makes sense that this factor could have an effect on group size.

So my project between now and next Spring is to disassemble 10,000 primers and weigh the three major components. Then load these primers into 10,000 identical rounds and test them at 1000 yards on a windless day.

Don't worry, when working on the primers I'll dial nine and one on my phone so that if anything goes wrong, I only have to dial one more number.

This will take some time, of course, but expect me to publish the results no later than April first, 2019.

Don't kill yourself...perhaps as few as 5000 would be sufficient .

 

[dbduff]

And you've actually seen raw data for a statistically relevant population? Good for you if you have. I haven't. However, I have seen this topic debated periodically on various internet shooting forums where somehow the relevant data always seems to be missing, which makes me a bit skeptical. I have no doubt that any primer manufacturer has more data on this than you or I could shake a stick at. Whether or not it's proprietary or is freely available is another story. Nonetheless, this thread has given me the motivation to test it for myself and I'll be happy to post the results here for others to draw their own conclusions and not simply assume that it works because someone else said so.

Believe you are commencing at the beginning of a very slippery slope. Good luck, we're pulling for you!
D

 

[dnellans]

Check this study out you may find it interesting to your own effort. I found the authors observations with Federal 210m mirror my own finding concerning ignitable primer mass I tested Wolf Large rifle and Fed 210s (approximately 30-35mg of a 350mg un-fired primer). The author also provides data concerning the relationship of total mass to peak pressure and ignitable mass to peak pressure. I am interested in your findings and methodology of testing if you can to share your results.

Link:
http://www.btgresearch.org/High-speed measurement of rifle primer blast waves.pdf

Thanks for putting data to the hearsay!

I wish the R value for primer weight to peak psi was a bit higher in that study (or more samples used) to make the data slam dunk, but it’s still pretty conclusive that there -is- enough variance in primers that you can both measure it and correlate it to unfired primer weight.

Is there any follow on study you know of that shows the primer variance will show up as reduced muzzle SD when igniting an actual powder/cartridge load?

Its also interesting that it definitely shows that changing primer types or using a magnum will have different characteristics. So playing with primers when in the final stages of load tuning should be considered as an option.

 

[DocBII]

dnellans,

Article would have been more useful if temperature of the gases was reported. It's also a factor.

DocBII

 

[dnellans]

dnellans,

Article would have been more useful if temperature of the gases was reported. It's also a factor.

DocBII

Do you have any other hard data you can add to the mix? I’d love to see anything quantitative, gas temp etc.

 

[CharlieNC]

This is the figure from the article which depicts the effect of primer weight on Peak Blast pressure. Frequently there are statements about statistical significance; using this as an example I will attempt to elaborate upon what are often misleading interpretations.



First of all your general impression is that there is some degree of cause-effect whereby higher primer mass leads to higher pressure. Secondly the scatter of the individual data points makes one question the effectiveness of sorting; for example fours shots with weight around 353 exhibited a pressure range of about 375-425, which is nearly the entire range across all the primer weights. But the common perception is that the correlation coefficient r=0.57 is "good". What does all this mean? A few highlights follow, and most if not all of these factors can be determined using Excel vs a statistical package.

Points:
a. Variability of the pressure is SD=30.9. The effort of the regression analysis is to explain as much of this SD as possible, using the simplest mathematical expression possible. While attempting to explain this with primer mass, the other implied aspects is that there is error involved measuring pressure and primer weight.

b. Is the correlation significantly significant? The correlation coefficient 0.57 means 0.57**2 = 30% r-squared of the pressure variability can be explained with this model. Additional parameters show there is a 93% chance this is significant.

c. But what about the scatter? The difference between measured pressure and that which is predicted by the regression line (residual = actual - predicted) has SD = 27.3, which provides information as to how well you will be able to predict pressure based on primer weight. In this case the SD of the error (27.3) is high vs the total pressure variability we started with (30.9).

d. How effective is sorting? Statistical packages such as Minitab include the ability to calculate another correlation coefficient r(predicted). This involves dropping an observation, calculating the predicted value of that observation using the remaining data, and then comparing that prediction to the actual; repeat for every observation. In this case r-squared(predicted) = 0.0%, meaning using weight to sort for pressure is completely ineffective.

So while there is a statistically significant correlation between weight and pressure, the lack-of-fit is so bad that it cannot be effectively used as a predictor. Why? Don't know exactly because what is lacking is information regarding the reproducability (precision) of pressure measurements and weighing, so we cannot begin to decifer whether the poor correlation is due to inaccuracies of one or both of these measurements, or if other significant factors are involved. The best correlation we could expect is that the SD of the residuals = SD of the precision of pressure measurement (unknown); which is another example of why just looking at the correlation coefficient "r" is a limited perspective.

So much for the lesson in statistics. The main point is that the original visual perspective of the graph is correct, some cause-effect going on but too much scatter to be useful. Don't always be swayed by analysis that contradict your common sense.

 

[Titan]

Thanks for putting data to the hearsay!

I wish the R value for primer weight to peak psi was a bit higher in that study (or more samples used) to make the data slam dunk, but it’s still pretty conclusive that there -is- enough variance in primers that you can both measure it and correlate it to unfired primer weight.

Is there any follow on study you know of that shows the primer variance will show up as reduced muzzle SD when igniting an actual powder/cartridge load?

Its also interesting that it definitely shows that changing primer types or using a magnum will have different characteristics. So playing with primers when in the final stages of load tuning should be considered as an option.

I haven't found a follow on to this particular study only a different one for 2014 that focus more time on the difference between lead free primers and standard primers. I will link that study below.

As to the relationship to muzzle velocity variance the total case capacity of the particular cartridge in question I believe matters. I saw more benefit in my limited testing in smaller cases like the .308 vs .284 winchester, however for my pupose (F Class) I saw a reduction in velocity variance with my lot of wolf primers and federal 210 so I sort by the milligram.


https://www.google.com/url?sa=t&sou...FjAAegQIARAB&usg=AOvVaw0esoYOhngaPsZeOZXGjGMd

 

[Ned Ludd]

This is the figure from the article which depicts the effect of primer weight on Peak Blast pressure. Frequently there are statements about statistical significance; using this as an example I will attempt to elaborate upon what are often misleading interpretations.

View attachment 1053602

First of all your general impression is that there is some degree of cause-effect whereby higher primer mass leads to higher pressure. Secondly the scatter of the individual data points makes one question the effectiveness of sorting; for example fours shots with weight around 353 exhibited a pressure range of about 375-425, which is nearly the entire range across all the primer weights. But the common perception is that the correlation coefficient r=0.57 is "good". What does all this mean? A few highlights follow, and most if not all of these factors can be determined using Excel vs a statistical package.

Points:
a. Variability of the pressure is SD=30.9. The effort of the regression analysis is to explain as much of this SD as possible, using the simplest mathematical expression possible. While attempting to explain this with primer mass, the other implied aspects is that there is error involved measuring pressure and primer weight.

b. Is the correlation significantly significant? The correlation coefficient 0.57 means 0.57**2 = 30% r-squared of the pressure variability can be explained with this model. Additional parameters show there is a 93% chance this is significant.

c. But what about the scatter? The difference between measured pressure and that which is predicted by the regression line (residual = actual - predicted) has SD = 27.3, which provides information as to how well you will be able to predict pressure based on primer weight. In this case the SD of the error (27.3) is high vs the total pressure variability we started with (30.9).

d. How effective is sorting? Statistical packages such as Minitab include the ability to calculate another correlation coefficient r(predicted). This involves dropping an observation, calculating the predicted value of that observation using the remaining data, and then comparing that prediction to the actual; repeat for every observation. In this case r-squared(predicted) = 0.0%, meaning using weight to sort for pressure is completely ineffective.

So while there is a statistically significant correlation between weight and pressure, the lack-of-fit is so bad that it cannot be effectively used as a predictor. Why? Don't know exactly because what is lacking is information regarding the reproducability (precision) of pressure measurements and weighing, so we cannot begin to decifer whether the poor correlation is due to inaccuracies of one or both of these measurements, or if other significant factors are involved. The best correlation we could expect is that the SD of the residuals = SD of the precision of pressure measurement (unknown); which is another example of why just looking at the correlation coefficient "r" is a limited perspective.

So much for the lesson in statistics. The main point is that the original visual perspective of the graph is correct, some cause-effect going on but too much scatter to be useful. Don't always be swayed by analysis that contradict your common sense.

Great analysis, Charlie. I also had a few issues with this work after reading it carefully, some of which you already addressed. Notably, the number of primers actually tested for the waveform analyses and tabular summary (Table 1) is given as "dozens". Presumably from Graph 3, n = 10, but we really have no definitive way of knowing the actual sample size (n) used for the comparisons of various brands of primers. This becomes an important point later in the work when the authors noted that sorting the primers to a "1 mg range" decreased the SD from 32.4 to 18.3, or approximately by half. Does that mean they only used n = 10 primers for both the waveform analyses and sorting tests? Or did the the SD decrease because "n" was smaller? This information was not directly stated in the report. The authors also did not provide any overall stats for the pool of 100 primers that were weight sorted into a "1 mg range" group. It would also have been useful to know the ES and SD for the weights of the entire 100 primer population in order to have a better feel for how narrow a weight-sorted "1 mg range" sample was when compared to the range of the population. If a "1 mg range" was a very small fraction of the overall population range, one might expect to have decreased SD more than by an approximate factor of 0.5.

In addition, the balance used to weigh primers in the subsequent tests (Figs. 3 and 4) was a torsion-type balance and was listed as having a resolution of 1 mg. In fact, only the readability of that balance is 1 mg, the actual resolution is almost certainly much lower, likely in the 2-3 mg range. That also may well explain the lack of fit to which you referred, because the primers were weight-sorted into 1 mg groups using a balance likely only capable of 2-3 mg precision under optimal conditions.

Nonetheless, it was gratifying to see an attempt made to quantify this approach, although the complete answer will likely require the results of a number of independent studies for a clear picture to emerge. This manuscript at least supports the notion that primer weight can be used as a surrogate measure for primer compound weight, a concept I find rather surprising. It will be interesting to determine how primer peak pressure as measured here correlates when using velocity as a readout, as I intend to do.

 

[retired]

charlie what long range bench rest discipline do you shoot 600 1000 ??

This is the figure from the article which depicts the effect of primer weight on Peak Blast pressure. Frequently there are statements about statistical significance; using this as an example I will attempt to elaborate upon what are often misleading interpretations.

View attachment 1053602

First of all your general impression is that there is some degree of cause-effect whereby higher primer mass leads to higher pressure. Secondly the scatter of the individual data points makes one question the effectiveness of sorting; for example fours shots with weight around 353 exhibited a pressure range of about 375-425, which is nearly the entire range across all the primer weights. But the common perception is that the correlation coefficient r=0.57 is "good". What does all this mean? A few highlights follow, and most if not all of these factors can be determined using Excel vs a statistical package.

Points:
a. Variability of the pressure is SD=30.9. The effort of the regression analysis is to explain as much of this SD as possible, using the simplest mathematical expression possible. While attempting to explain this with primer mass, the other implied aspects is that there is error involved measuring pressure and primer weight.

b. Is the correlation significantly significant? The correlation coefficient 0.57 means 0.57**2 = 30% r-squared of the pressure variability can be explained with this model. Additional parameters show there is a 93% chance this is significant.

c. But what about the scatter? The difference between measured pressure and that which is predicted by the regression line (residual = actual - predicted) has SD = 27.3, which provides information as to how well you will be able to predict pressure based on primer weight. In this case the SD of the error (27.3) is high vs the total pressure variability we started with (30.9).

d. How effective is sorting? Statistical packages such as Minitab include the ability to calculate another correlation coefficient r(predicted). This involves dropping an observation, calculating the predicted value of that observation using the remaining data, and then comparing that prediction to the actual; repeat for every observation. In this case r-squared(predicted) = 0.0%, meaning using weight to sort for pressure is completely ineffective.

So while there is a statistically significant correlation between weight and pressure, the lack-of-fit is so bad that it cannot be effectively used as a predictor. Why? Don't know exactly because what is lacking is information regarding the reproducability (precision) of pressure measurements and weighing, so we cannot begin to decifer whether the poor correlation is due to inaccuracies of one or both of these measurements, or if other significant factors are involved. The best correlation we could expect is that the SD of the residuals = SD of the precision of pressure measurement (unknown); which is another example of why just looking at the correlation coefficient "r" is a limited perspective.

So much for the lesson in statistics. The main point is that the original visual perspective of the graph is correct, some cause-effect going on but too much scatter to be useful. Don't always be swayed by analysis that contradict your common sense.

 

[butchlambert]

charlie what long range bench rest discipline do you shoot 600 1000 ??

I guess I need to spend about 10 grand to build a suitable LR rifle, scope, and reloading tools to have competitive equip. to shoot another discipline. At a minimum I will need a couple grand additional to get to Montana, rooms, food, and entry fees.
I can use that money to start on another walnut stocked custom rifle. I doubt that you will find me posting about buying my products, except on the forum for the paying advertisers.
How about you bringing yourself down this way and shoot shortrange BR with us?

 

[CharlieNC]

charlie what long range bench rest discipline do you shoot 600 1000 ??

Mainly midrange Fclass, shooting a 223. A few "tactical" steel out to 1200yd.

 

[retired]

well i guess that covers it, opinions on a discipline you do not shoot.
my most expensive rifle cost 6k in parts. my heavy is cheap 350 stock, 100 trigger, 700 action 350 bbl. and the scope was all of 1000.
sounds like excuses to me.
maybe you should keep your opinions where your experience is.
this subject is all long range.

I guess I need to spend about 10 grand to build a suitable LR rifle, scope, and reloading tools to have competitive equip. to shoot another discipline. At a minimum I will need a couple grand additional to get to Montana, rooms, food, and entry fees.
I can use that money to start on another walnut stocked custom rifle. I doubt that you will find me posting about buying my products, except on the forum for the paying advertisers.
How about you bringing yourself down this way and shoot shortrange BR with us?

 

[retired]

again an opinion with no application. bench rest, long range.

Mainly midrange Fclass, shooting a 223. A few "tactical" steel out to 1200yd.

 

[dolek]

I smell alcohol...

 

[butchlambert]

well i guess that covers it, opinions on a discipline you do not shoot.
my most expensive rifle cost 6k in parts. my heavy is cheap 350 stock, 100 trigger, 700 action 350 bbl. and the scope was all of 1000.
sounds like excuses to me.
maybe you should keep your opinions where your experience is.
this subject is all long range.

Kid, I don't shoot with junk. Why would you use less than the best? Components are only important in LR?
Post photos of your wood after the Montana matches.

 

[Dos XX]

My first use of the Ignore button. Kind of felt good, too.

 

[BoydAllen]

First of all you will need to have a way to accurately weigh charges to a very small variation...to minimize that variable. Then you could do a test, loading at the range, with a single case. Weigh a hundred primers and take five of the heaviest and five of the lightest. Chronograph the velocities. Absent positive compensation, differences in trajectory can be calculated from velocities. Looking at targets throws in tune, adding a variable rather than clarifying the results. If there is a significant difference in the average velocity of the heavy group compared to the light group then it would seem that for uses where small differences in velocity are thought to be important that sorting primers by weight would be a good idea.

 

[CharlieNC]

again an opinion with no application. bench rest, long range.

Do you have a reading comprension problem? ALL of my discussion dealt strictly with a detailed review of ONE graph, and the frequent misinterpretation of it and similar content. Nothing more, and nothing less.

 

[dmoran]

My own experience; I've had primer lot's that weight sorting made no difference. Lot's that weighing made small differences. And lot's that weight sorting made fairly substantial differences.
Personally don't care what any statistical data may yield to weight sorting primers. What matters to me is what my targets yield, and what each primer lot tests and proves out to be. It is part of my reloading program for competition and always will be, because it has repeatedly proven out to me.

 

[seymour fish]

Believe you are commencing at the beginning of a very slippery slope. Good luck, we're pulling for you!
D

Seems he has taken his own bait